QTGMC is slower for me with 10-bit video than 8-bit video, anyone else?

[ENunn]

running my ld-decode ffv1 10-bit rips through avisynth and my script is really really slow. im getting about 3fps with a deinterlace, aspect correction, crop, upscale, and prefetch. it's a powerful script with several options enabled with qtgmc, but the same script with the video converted to 8 bit before qtgmc gives me about 4.85fps. transcoding the source file to prores speeds it up slightly but it's still slow. my vhs scripts with 8-bit avi source files and a lot more filters run at the same speed as 10-bit ffv1+deinterlace+aspect correction+crop+upscale+prefetch.

i also ran the video with lighter qtgmc settings, namely preset=faster, and with the video at 10-bit i get 32fps and when i convert it to 8-bit i get 56fps

is it my qtgmc version or is it like this for everybody? i'm running avisynth+ 3.7.4 and qtgmc 3.386.

scripts:

#BASIC
v = lwlibavvideoSource("d:\Domesday LD Rips\the agony and the ecstasy 1989 laserdisc opening-corrected.mkv") 
a = lwlibavaudiosource("d:\Domesday LD Rips\the agony and the ecstasy 1989 laserdisc opening-corrected.mkv")
audiodub(v,a)
assumetff()
Trim(30, 904)
Trim(0, 510)
qtgmc(preset="faster")
converttoyuv420(matrix="rec709", interlaced=false)
prefetch(6)

#ADVANCED/FOR YOUTUBE
v = lwlibavvideoSource("d:\Domesday LD Rips\the agony and the ecstasy 1989 laserdisc opening-corrected.mkv") 
a = lwlibavaudiosource("d:\Domesday LD Rips\the agony and the ecstasy 1989 laserdisc opening-corrected.mkv")
audiodub(v,a)
assumetff()
Trim(30, 904)
Trim(0, 510)
QTGMC(TR0=2, TR1=2, TR2=2, Rep0=1, Rep1=0, Rep2=4,
  \ DCT=5, ThSCD1=300, ThSCD2=110,
  \ SourceMatch=3, Lossless=2, Sharpness=0.0, SLMode=0, Sbb=0, MatchPreset="slow",
  \ NoiseProcess=2, GrainRestore=0.0, NoiseRestore=0.4, NoisePreset="slow",
  \ StabilizeNoise=false, NoiseTR=0, NoiseDeint="generate")
converttoyuv444(matrix="rec601", interlaced=false)
  resizeshader(651,488,kernel="ssim")
SuperRes(matrixin="rec601", 2, .43, 0, """nnedi3_rpow2(4, nns=4, cshift="Spline36resize")""", matrixout="rec709")
crop(32,24,-24,-12)
x = float(width) / float(height)
resizeshader(width=round(x * 2160),height=2160, kernel="ssim")
converttoyuv420(matrix="rec709", interlaced=false)
prefetch(6)

[Selur]

Isn't that to be expected, since 10bit has more data to process

Format              Luma (Y)   Chroma (U & V) Chroma Subsampling Bit Depth  Storage per sample
YUV420 8-bit     1 byte       1 byte each       4:2:0 (quarter res)     8-bit          1 byte (8 bits)
YUV420 10-bit   2 bytes      2 bytes each     4:2:0 (quarter res)     10-bit         2 bytes (16 bits)

?

[ENunn]

Isn't that to be expected, since 10bit has more data to process

Format              Luma (Y)   Chroma (U & V) Chroma Subsampling Bit Depth  Storage per sample
YUV420 8-bit     1 byte       1 byte each       4:2:0 (quarter res)     8-bit          1 byte (8 bits)
YUV420 10-bit   2 bytes      2 bytes each     4:2:0 (quarter res)     10-bit         2 bytes (16 bits)

?

probably, I just wanna make sure if this is normal or not

[Selur]

It is normal that using 10bit is slower.